Which Airbnb Property Is Losing You Money

Two corrections re-rank most portfolios: Airbnb's fee is 15.5% now, not 3%, and splitting shared costs evenly moves about $870 a year onto your smallest door.

The property losing you money is usually not the one your spreadsheet names. Two corrections re-rank most portfolios: Airbnb's host service fee is now 15.5% rather than 3%, which is 12.5 points of gross revenue, and splitting shared costs evenly across doors loads roughly $870 a year onto your smallest property that a revenue-weighted split would not.

On a six-door portfolio grossing $305,000 a year, modelling the retired 3% fee understates Airbnb's cut by $38,125. That single error is larger than the annual profit of most individual doors. It lands hardest on the smallest one, where correcting the fee adds $3,437.50 of cost against $27,500 of revenue.

Which Airbnb property is losing you money changes the moment you fix the fee line

Fix the fee before you rank anything. Airbnb now charges hosts a single service fee, and its help centre states the rate plainly: most hosts pay 15.5%, remaining hosts typically pay 14% to 16%, and listings in Brazil and Mexico pay 16%. The old structure took 3% from the host and 14.1% to 16.5% from the guest. Airbnb's own worked example is the cleanest way to see the difference: keep your nightly price at $100 and you now receive $84.50.

The deadline to reprice passed on 15 September 2026 for hosts outside the European Economic Area, and falls on 13 October 2026 inside it. As of 26 September 2026, Airbnb has published no page describing what happens after the first deadline; every fee page still writes it in the future tense. The rate itself is on Airbnb's service fees help article, read 26 September 2026.

Almost nothing online has caught up. Across twelve pages ranking for the questions an operator actually types about Airbnb profitability, not one models 15.5% as the primary figure. Seven still model 3% or 3% to 5% with no indication anything changed, and one of those carries an update stamp of 18 September 2026, three days after the non-EEA deadline. Airbnb's own host-education page on what it charges hosts still leads with the split fee and the 3% line. A recent update date is not evidence the fee was corrected.

Recompute every door before you conclude anything about any of them. MagicBNB's Property Health Grid derives each card's health dot from margin rather than revenue, so the ranking re-orders itself when a cost line changes across the portfolio. On the composite above, correcting the fee moves two doors out of green, and the door that had looked worst on revenue is not the one that turns red.

Three reasons a door looks unprofitable, and only one of them is the door

A property showing a loss is reporting one of three different things, and the fix differs for each.

• It genuinely loses money. Revenue does not cover the costs that would disappear if the property did. This is the only case where selling or exiting is the answer.
• It is mis-allocated. Portfolio-level costs were dumped on it, or a cost that belongs to it was quietly absorbed somewhere else. The property is fine; the split is arbitrary.
• It is measured against the wrong revenue. The number on the books is Airbnb's gross reservation total rather than the money that reached the bank, and those differ by more than most operators expect.

The third case has a published worked example. Airbnb's US tax documents page shows a single reservation reported at $600 on the Form 1099-K (a $500 nightly total, a $90 cleaning fee, $10 of pass-through tax) against a net payout of $467.60 after $18 of Airbnb fees and $114.40 of co-host payouts. That is a gap of $132.40, or 22.07% of the reported figure, on one booking. Source: Airbnb, US tax documents from Airbnb, read 26 September 2026.

Read the fee line in that example. $18 against a $590 pre-tax subtotal is 3.05%, so Airbnb's own 1099-K illustration is still built on the split fee it is retiring, and Airbnb publishes no single-fee version of it. If you reconcile your books against that page, you inherit the old rate twice over.

The IRS does not tell you how to split a cost between two doors

There is no prescribed method, and that absence is the whole problem. Publication 527 carries three allocation rules: rental versus nonrental travel, land versus building for depreciation basis, and rental versus personal use of one dwelling. Only the last prescribes a method, and its standard is permissive: you can use any reasonable method for dividing the expense, with rooms or square footage offered as the common approaches. Nothing in Publication 527, the Schedule E instructions or the form itself tells you how to divide one bill across three properties.

Schedule E gives each property its own column, A, B and C, with expense lines 5 through 19 repeated in each. There is no shared-cost column. Three properties fit on one form, and with more than three you attach additional Schedules E and complete the Totals column on only one of them. The recordkeeping standard is on the IRS page for rental real estate income, deductions and recordkeeping, last reviewed 4 April 2026, and it requires documentary evidence such as receipts, cancelled checks or bills. Substantiating the expense is the whole obligation. Putting it on the right door is not.

Schedule E then nets everything anyway. Line 24 adds the positive amounts, line 25 adds the losses, line 26 combines them. A bleeding door and a strong one produce one number, and the return foots whichever way you split the overhead. Your tax filing will never tell you which property is the problem.

Nothing in the tax code forces the per-property truth. The return foots either way, which is exactly why the number you manage from has to be built for a different purpose than the number you file.

Take a six-door Phoenix-area composite grossing $305,000 across doors of $71,400, $63,200, $58,900, $49,200, $34,800 and $27,500. Its portfolio-level costs run $11,400 a year and none arrives as a per-door invoice: $1,428 for Hospitable Professional at six properties, $1,680 for QuickBooks Online Plus (the cheapest tier with class tracking), $400 for an AirDNA Market Research seat, $3,118.50 of vehicle mileage and a $4,773.50 master insurance premium covering all six units.

Split evenly, that is $1,900 a door. Split by revenue share, the smallest door carries $1,027.87 and the largest carries $2,668.72. The smallest door's bill moves by $872.13 depending on which method you picked, and you picked it arbitrarily, because nobody made you pick either one.

On that door the swing decides the verdict. It contributes $1,397.50 after its own direct costs. Under the even split it reports a loss of $502.50. Under revenue share it reports a profit of $369.63. Same property, same year, same bank statements.

The mileage line shows how fine the error gets. The IRS published two business standard mileage rates for 2026: 72.5 cents through 30 June, 76 cents from 1 July. Run 4,200 miles evenly across the year and the deduction is $3,118.50. Use one rate for the whole year and you get $3,045 or $3,192, then split a wrong total six ways.

Pick one allocation method and make it hold without you. MagicBNB's Recurring rules tie a merchant to a property split once, then apply it to every future transaction from that merchant and backfill the past ones. The utility bill, the insurance debit and the software subscription stop being re-decided each month, which is what makes a year-over-year comparison between two doors mean anything.

Measure each door against its own market's percentile spread

A national average tells you almost nothing about one address. AirROI's February 2026 market data publishes RevPAR percentiles per market, and the spread inside a single market is enormous: in Phoenix the 25th percentile listing runs $57 a night of RevPAR, the median $126, the 75th percentile $239 and the 90th percentile $437. In Las Vegas the bottom quartile earns $27 a night against $219 at the top decile, an 8.0x gap. AirROI's own conclusion from that data is that in every oversaturated market it analysed, the bottom quartile generates revenue insufficient to cover standard carrying costs.

Put the composite's smallest door against that scale. $27,500 across 365 available nights is a RevPAR of $75, which sits between the Phoenix 25th and 50th percentiles and much nearer the bottom. Its largest door runs $196, above the median and below the 75th percentile. Percentiles read 26 September 2026 from AirROI's market oversaturation analysis, which publishes its own February 2026 percentile cuts.

Ratings explain more of that spread than most operators credit. On 7,203 Nashville listings in March 2026, AirROI measured trailing-twelve-month revenue of $42,935 for listings rated 4.90 or above, $38,676 at 4.80 to 4.89, $33,669 at 4.70 to 4.79, $21,914 at 4.50 to 4.69, and $11,833 below 4.50. The steps are not even. Falling from the 4.70s into the 4.60s costs $11,755, roughly triple the cost of the first step down from 4.90. A door sitting at 4.68 is not a marginally worse property than one at 4.72.

One warning on the cost side of any such comparison. No primary source publishes a short-term rental expense ratio that was actually measured. AirROI asserts 40% to 60% of gross revenue on one glossary page and an NOI margin on another that does not reconcile with it, and neither is derived from observed property expenses. Borrow that number and you are benchmarking against an editorial assumption.

Run the marginal test before you decide anything

A door's reported loss is the wrong input for an exit decision. Price the exit on what actually disappears when the door does. Take the composite's smallest property again. Selling it removes $27,500 of revenue and $26,102.50 of direct cost, so the portfolio loses $1,397.50 of contribution. Very little overhead goes with it: the Hospitable per-property increment of $15 a month, or $180 a year, and whatever the carrier takes off the master premium for dropping one unit, which on this composite was $795. Total avoided, $975.

So the exit costs $422.50 a year, not the $502.50 it appeared to save. And the remaining $10,425 of portfolio cost now spreads across five doors instead of six, which is $2,085 each under an even split, up $185. On this portfolio that is enough to push the next-smallest door, contributing $2,010, from a reported $110 profit into a reported $75 loss. Even-split allocation makes exits contagious: remove the door at the bottom and it manufactures a new one.

Revenue-weighted allocation does not do this, because the base shrinks with the portfolio. That is a real argument for the method, and it is a better one than any appeal to fairness.

What to do with a door that is genuinely losing money

Treat it as three candidate decisions and price each one, rather than picking the first that feels right. Reprice, meaning reset the base rate and the length-of-stay rules against current market percentiles. Fix the operational cause, most often a rating sitting just under a revenue step. Or exit, which only wins when the avoidable-cost saving beats the lost contribution.

The rating case usually has the best arithmetic, because the cost of the gap is measurable and the fix is cheap. Moving a door from the 4.60s into the 4.70s is a cleaning and response-time problem far more often than a capital one, and on the Nashville data that step is worth $11,755 a year.

When the three options are close, model them rather than argue about them. MagicBNB's Milo analyst runs Tree-of-Thoughts scenarios for exactly this decision shape, generating a renovate, reprice and hold path, scoring each on revenue, ROI timeline, risk and long-term implications, then recommending one with the reasoning attached. Keep that reasoning. A door held on an argument you cannot reconstruct in six months is a door you re-litigate in six months.

One caveat on all of this. No published study measures the spread of performance inside a single operator's portfolio. The nearest peer-reviewed evidence, Törnberg's 2022 study of 97 Airbnb markets in PLoS ONE, found an average Gini coefficient of 0.68 across hosts, with 10% of hosts taking roughly half of market revenue. That is inequality between hosts, on 2019 revenue. Market percentile is the best available stand-in for where one of your doors sits, and it should be used as one.

Frequently asked questions

How do I know which Airbnb property is losing money?

Compare each door's revenue against only the costs that would disappear if you no longer owned or leased it, then allocate portfolio-level costs by revenue share rather than evenly. On a six-door portfolio with $11,400 of shared cost, the two methods differ by $872 a year on the smallest door, which is enough to flip its reported sign. Before running either, recompute Airbnb's fee at 15.5% rather than 3%.

Should I sell an Airbnb property that shows a loss?

Only if the costs you would actually avoid exceed the contribution you would lose, which is rarely what a per-door P&L shows. On the composite in this post, a door reporting a $502.50 loss was contributing $1,397.50 against $975 of genuinely avoidable cost, so selling it made the portfolio $422.50 a year worse and pushed the next door into a reported loss.

How should I split shared costs between rental properties?

Revenue share is the better default, because the allocation shrinks with the portfolio when a property leaves and does not manufacture a new loss-maker. Even splits load small doors disproportionately: on an $11,400 annual pool across six doors, the smallest property carries $1,900 under an even split and $1,027.87 under revenue share.

Does the IRS require per-property profit tracking?

No. The IRS requires you to substantiate each expense with documentary evidence, not to assign it to the correct property, and it prescribes no method for splitting a cost across multiple rentals. Schedule E gives three property columns and nets all of them at lines 24 through 26, so a return with a hidden loss-maker foots exactly as well as one without.

Why does my Airbnb 1099-K not match my bank deposits?

The 1099-K reports gross reservation totals before Airbnb fees and before co-host payouts, so the gap is structural rather than an error. Airbnb's own published example shows $600 reported against a $467.60 payout, a 22.07% difference on a single booking. The reporting threshold is currently gross above $20,000 and more than 200 transactions, with nine states filing lower.

What is a normal expense ratio for a short-term rental?

No primary source publishes a measured one, so treat every quoted figure as an assumption. The ranges in circulation, including AirROI's 40% to 60%, are asserted rather than derived from observed property expenses. Build the ratio from your own reconciled costs and compare each door against your portfolio rather than a borrowed benchmark.

Key takeaways

• Airbnb's host service fee is 15.5% for most hosts, not 3%. On a portfolio grossing $305,000 a year, modelling the retired rate understates the fee by $38,125.
• Across twelve pages ranking for Airbnb profitability questions, zero model 15.5% as the primary figure and seven still model 3%, including one updated 18 September 2026.
• Splitting $11,400 of annual portfolio cost evenly across six doors charges the smallest one $1,900; revenue share charges it $1,027.87, a $872.13 swing that flips its reported sign.
• The IRS prescribes no method for allocating a cost across multiple rental properties, and Schedule E nets every property at lines 24 through 26, so a tax return never identifies the loss-maker.
• Phoenix RevPAR ran $57 at the 25th percentile and $437 at the 90th in February 2026, so a door has to be measured against its market's distribution rather than a national average.
• On 7,203 Nashville listings in March 2026, falling from a 4.70-4.79 rating into 4.50-4.69 cost $11,755 of trailing-twelve-month revenue, roughly triple the first step down from 4.90.

Read next

The category structure that makes a per-property split possible in the first place is in STR Expense Tracking: The Complete Guide, and the line-by-line read of the statement those categories produce is in How to Read an STR Profit and Loss Statement.

For the occupancy a specific door has to hit before it contributes anything, see STR Break-Even Occupancy Rate. For the cleaning-fee ratio that moves both occupancy and revenue, see Airbnb Cleaning Fee Strategy.

Your books balance. Your Schedule E foots. Neither one tells you which of your six doors is carrying the other five. See per-door profit in MagicBNB →

About MagicBNB

MagicBNB is the portfolio analytics layer for operators running 2 to 20 short-term rental doors. The Smart transaction ledger reads every bank transaction, suggests a category with a confidence band, and opens a multi-split dialog when one payment belongs to more than one property, which is where most allocation errors are actually made. Discovery spotlights name the patterns the numbers imply, flagging a door as a cash cow, a grinder, a silent winner or a cleaning burden rather than leaving you to spot it. The Deal Analyzer keeps every saved analysis side by side and ranks them against your own risk tolerance and target return, so an exit decision is compared rather than argued. See what your portfolio earns at magicbnb.io.